Unweave

ML, without the FML

Hey Product Hunt!

My name is Noorvir and I’m the founder of Unweave. Together with @mark_e_winter and @matthewmihok we’re building the Heroku for Machine Learning.

Unweave gives you instant serverless infrastructure to train your machine learning models. It gets you up and running with fully reproducible training environments with near zero config.

The inspiration for Unweave, and the tagline, came from the countless hours spent spinning up EC2 instances, managing S3 buckets, tweaking config files and writing heaps of boilerplate code just to get some extra processing power and storage capacity. It led us to ask ourselves: what if you could keep that gleamingly smooth Macbook Pro user experience but bundle it with infinite storage and computing capacity?

That’s what we’ve set out to build with Unweave.

Unweave works in one simple way: if you run `unweave python train.py` instead of `python train.py`, it automatically builds and runs your environment in the cloud, within seconds, while versioning everything you run end-to-end.

This is a huge unlock for any engineer or company trying to explore and learn from their data. The ability to take a script and run it instantly in the cloud while feeling like it's still local is an unlock similar to what cloud dev environments, and some of what Github Codespaces are enabling.

Nice work @noorvir!